Daily OC on i7 13700k became unstable after BIOS update

[S3se]

Hi everyone. At the end of February I tweaked a small OC to my i7 13700k. Just 55x to p-cores and 44x to e-cores. Also I enabled xmp 1 profile and did a small undervolt with -0.01V on adaptive voltage. Everything was stable and I could rerun Cinebench r23 again and again. After BIOS update the same settings became unstable and Cinebench keep crashing just on the start of multicore test. Few days ago I updated BIOS to the newest version 0904 and still the same thing. Could somebody answer me what could be a possible reason?
My setup:
Cpu: i7 13700k
Motherboard: asus rog strix z790-f wifi
Ram: G.Skill DDR5 32GB (2x16GB) 6000 Mhz Trident Z5 RGB Black (F5-6000J3636F16GX2-TZ5RK)

[Nate152]

Hello S3se 

What I'd do is, reset the bios to defaults, enable X.M.P. and try cinebench.

If it passes, then your mild cpu overclock isn't quite stable.

[S3se]

Hi Nate152

Yeah, i tried different configurations. With all default settings in bios cinebench r23 stress test passes normaly. Then I enabled xmp and also cinebench passed normally. After that i decided to set 54x on p-cores and 43x on e-cores with llc level 4 and it passed the test too. 

The very one really confusing thing that even if i set 55x and 44x and set llc to 5, cinebench just crashes after a few seconds. Voltage on cores during those few seconds of test was 1.280-1.289. Meanwhile on some oc videos where guys set 56x on p cores and 44x on e cores, their 13700k passed cinebench with lower voltage. Their average was ~1.272. So, I'm still confused. Am I really unlucky with silicon?

[Nate152]

Are you syncing all cores to 55x and 44x or are you doing per core?

If you're syncing all cores, you'll need more voltage.

[S3se]

Yes, I sync to all cores. Recently I tried with +0.2 offset on adaptive voltage, during some seconds before crashing I saw voltage between 1.295-1.310. As I mentioned in first post, I could set 55x and 44x sync all cores before updating to new versions of BIOS. As I remember, average voltage on cores was 1.278. But for now even with plus offset and higher voltage as a result, system can't pass cinebench test.

[Nate152]

In this case, you can roll back to the bios version that was more stable for you and wait for future updates.

[S3se]

Thanks for answer. That's not something critical for me, just my curiosity and experimenting. But still it's strange behaviour. I don't know what could be a reason and I can't do even a small oc. I'll keep waiting for new updates

[S3se]

Ok, now I'm completely upset. I really don't know what is going on. I downgraded to the BIOS version I had when I firstly launched this pc, version is 0703 and the same thing happens with cinebench! Set 55x sync all cores on p-cores and just crashes everytime even with 1.297V on cores. Tried freshly installed cinebench r23 and that didn't help

[Nate152]

When overclocking, it's probably better to enter the voltage manually rather than just using offsets.

Try entering in 1.31v with LLC at level 6 then try Cinebench. Monitor the maximum voltage, it should be at or close to 1.31v.

If the maximum voltage is higher than 1.31v, lower LLC a step.

If the maximum voltage is lower than 1.31v, raise LLC a step.

If it's still crashing, give it a little more voltage up to 1.35v and see if it passes. Only overclock the P-cores for now, get the P-cores stable first then overclock the E-cores.

[S3se]

Thanks, previously I only did oc with adaptive voltage. I will try your advise and reply soon